Hostile TME, especially in breast cancer, is characterized by enriched FAs and other lipid species, which foster immunosuppression and support tumor growth and metastasis (Xu et al., 2021; Zhang et al., 2017b). Since palmitic acid (PA, 16:0), stearic acid (SA, 18:0), oleic acid (OA, 18:1), and linoleic acid (LA, 18:2) are the most common FAs in humans, as shown in Figure 1A, we treated macrophages with individual albumin-conjugated FAs and determined their fate. Using a macrophage cell line, we demonstrated that all the FAs studied were oxidized by mitochondria as evidenced by the increased oxygen consumption rate (OCR) when compared to the bovine serum albumin (BSA) control (Figure 1B). Interestingly, when measuring their impact on other cellular organelles, FAs exhibited dichotomous effects, with saturated FAs increasing lysosome contents (Figure 1C), while unsaturated FAs promoting ER (Figure 1D) and, more dramatically, lipid droplet (LD) formation (Figure 1E) in macrophages. Using primary peritoneal macrophages (Figure 1—figure supplement 1A) and bone-marrow-derived macrophages (Figure 1—figure supplement 1B), we observed the similar phenomenon that unlike saturated FA, unsaturated FA significantly induced LD formation in these macrophages. Oil Red O staining confirmed the presence of LD formation (Figure 1—figure supplement 1C). Of note, the observation of unsaturated FA-induced LD formation in macrophages was consistent regardless of FA concentrations (Figure 1—figure supplement 1D) and treatment durations (Figure 1—figure supplement 1E). With a multispectral imaging analysis, we demonstrated that OA/LA-induced LD accumulation was not associated with the lysosome, but colocalized in the ER (Figure 1F). Transmission electron microscope further confirmed that OA and LA specifically induced LD formation in the cytoplasm (Figure 1G), which were bud from ER (Figure 1—figure supplement 1F). Collectively, these data suggest that unsaturated FA-induced LD biogenesis in macrophages mainly occurs in the ER.

It is well known that key enzymes, including GPATs, AGPATs, LIPINs, DGATs, are involved in TAG synthesis and LD biogenesis (Figure 2A). To determine how unsaturated FAs, but not saturated FAs, induced LD biogenesis in macrophages, we treated macrophages with either LA or PA, and measured their impact on LD biogenesis-related enzymes. In response to PA or LA treatment, Acs1, Cpt1a, Cpt1b, HMGCR, Acat1, and Acat2 had similar expression pattern in macrophages (Figure 2—figure supplement 1A–F), suggesting that both FA types exerted similar effects on FA activation, mitochondrial oxidation, cholesterol synthesis and esterification. Interestingly, genes encoding rate-limiting enzymes in TAG synthesis, including Gpam1, Dgat1 and Dgat2, were significantly increased by treatment with LA, but not PA (Figure 2B–H), suggesting a unique role of LA in transcriptional upregulation of neutral lipid synthesis pathways. We observed by confocal microscopy that macrophages treated with LA, but not BSA or PA, exhibited elevated levels of LipidTOX, GPAT1, and DGAT1 (Figure 2I). Quantitative measurements showed a significant increase of lipid accumulation (LipidTOX), GPAT1 and DGAT1 proteins in the LA-treated vs. the BSA or PA groups (Figure 2J–M). Moreover, intracellular flow cytometric staining (Figure 2—figure supplement 1G–I) and quantification by mean fluorescent intensity (MFI) (Figure 2—figure supplement 1J–L) also confirmed our observations that unsaturated FAs significantly increased lipid accumulation by enhancing GPAT/DGAT-mediated lipid accumulation in macrophages.

To determine how unsaturated LA transcriptionally upregulated gene expression in the TAG synthesis pathway, we searched for transcription factor binding sites, focusing on the pooled genes related to TAG synthesis pathways. Two transcriptional factors, C/EBPα and C/EBPβ, were predicted to commonly bind the promotor regions of these genes (Figure 3—figure supplement 1A). When macrophages were treated with PA or LA, C/EBPα (Figure 3A), but not C/EBPβ (Figure 3—figure supplement 1B), was significantly upregulated by LA treatment. Confocal microscopy showed that C/EBPα was mainly localized in the cytosol with BSA and PA treatment but was significantly upregulated in the nuclei with LA treatment (Figure 3B and C). In contrast, C/EBPβ was present in both cytosol and nuclei regardless of PA or LA treatment (Figure 3—figure supplement 1C and D). These results suggest that LA treatment induces specific activation of C/EBPα in macrophages.

To further dissect the role of C/EBPs in LA-treated macrophages, we knocked down C/EBPα (Figure 3D) and C/EBPβ (Figure 3—figure supplement 1E), respectively, in macrophages. LA-induced gene expression of Gpam1, Dgat1, Dgat2 was abrogated when C/EBPα was silenced (Figure 3E–G). Moreover, the uptake of FA (Cd36), lysosome lipase (Lipa) and mitochondrial oxidation (Cpt1b) were significantly increased in C/EBPα-knockdown macrophages (Figure 3H–J), suggesting that C/EBPα knockdown not only inhibited lipid synthesis, but also promoted lipolysis and oxidation. As such, C/EBPα knockdown significantly inhibited LA-induced neutral lipid accumulation as shown by the staining of BODIPY (Figure 3K) and Oil Red O (Figure 3L and M) in macrophages. By contract, C/EBPβ silencing did not affect the expression of Gpam1, Cd36, and Lipa expression (Figure 3—figure supplement 1F–H). Although C/EBPβ knockdown appeared to inhibit the expression of Dgat1 and Dgat2, it also inhibited Cpt1b-mediated FA oxidation (Figure 3—figure supplement 1I–K). Overall, C/EBPβ knockdown resulted in lipid accumulation in response to LA treatment in macrophages (Figure 3—figure supplement 1L). Altogether, these data suggest that C/EBPα, but not C/EBPβ, is a critical transcriptional factor controlling LA-induced TAG synthesis and LD formation in macrophages.

Given the insolubility of long chain FAs, it was of great interest to dissect how LA, unlike SA, specifically activated C/EBPα in macrophages. As FA chaperones, FABPs bind different FAs, facilitating their transport and responses. We measured the profile of common FABP members in macrophages and demonstrated that FA treatment mainly induced expression of FABP4, but not FABP3 and FABP5, in macrophages (Figure 4A–C). Analyzing the TCGA human breast cancer database, we found that compared to other FABP family members FABP4 was mostly correlated with C/EPBα (Figure 4—figure supplement 1A and B), but not with other C/EBP members (Figure 4—figure supplement 1C). The positive associations of FABP4 with C/EBPα and key TAG synthesis genes were also evidenced in multiple GEO datasets (Figure 4—figure supplement 1D–F).

To gain insights into our observations, we performed single cell RNA-sequencing using splenic macrophages. UMAP plots showed that macrophages consisted of 13 clusters based on untargeted gene expression profiles (Figure 4—figure supplement 1G). FABP4 was enriched in clusters 0, 1, and 2 (Figure 4D). Compared to other non-enriched clusters (Fabp4-), GO molecular function analysis showed that FABP4 was mainly involved in protein-lipid complex, lipid binding and signaling receptors (Figure 4—figure supplement 1H). Moreover, Fabp4 enriched clusters (Fabp4+) were highly associated with genes known to be critical for lipid uptake and accumulation, including Cd36, PPARγ, CEBPA, Plin2 (Figure 4E–I). In contrast, Fabp4 enrichment was not related to Fabp5, Acsl1, Lipa, Fasn, and Cpt1b (Figure 4J, Figure 4—figure supplement 1I–L), suggesting a unique role of FABP4 in lipid uptake and storage in macrophages.

Confocal microscopy showed that FABP4 was mainly expressed in the cytosol in BSA- or PA-treated cells, whereas LA treatment led to FABP4 present both in the cytosol and nuclei (Figure 4K), suggesting a LA-induced FABP4 nuclear translocation effect for potential nuclear transactivation. To confirm the role of FABP4, we measured C/EBPα expression in FABP4-deficienct macrophages. Deletion of FABP4 in macrophages abrogated LA-induced C/EBPα expression in nuclei (Figure 4L and M). We further confirmed that FABP4 deficiency significantly blocked LA-induced expression of C/EBPα (Figure 4—figure supplement 1M and N), but not C/EBPβ (Figure 4—figure supplement 1O and P), in primary peritoneal macrophages and bone-marrow-derived macrophages (BMMs). Collectively, these data demonstrated a pivotal role for FABP4 in mediating LA-induced C/EBPα activation in macrophages.

Given the critical role of FABP4 in activating the LA-C/EBPα axis, we reasoned that FABP4 deficiency suppressed LA-induced lipid accumulation in macrophages. To this end, we first measured if FABP4 deficiency reduced key enzymes of TAG synthesis in different macrophages. Using FABP4 wild type (WT) and knockout (KO) macrophage cell lines (Figure 5A), we demonstrated that upregulation of Gpam1, Dgat1, and Dgat2, but not Gpam4 and Agpat2, in response to LA treatment in WT macrophages, was significantly reduced when FABP4 was genetically depleted (Figure 5B–F). Similarly, using peritoneal macrophages purified from WT and KO mice, we confirmed that FABP4 deficiency significantly reduced the expression of LA-induced Gpam1, Dgat1, Dgat2 (Figure 5—figure supplement 1A–D). Of note, FABP4 deficiency also reduced expression of other lipid transporters, including Cd36 and Fabp5 (Figure 5—figure supplement 1E–F), suggesting a general inhibition of fatty acid uptake and metabolism in the absence of FABP4 in primary macrophages.

To evaluate the role of FABP4 in lipid accumulation, we measured GPAT1 and DGAT1 protein expression and lipid droplet formation in LA-treated WT and KO macrophages. Confocal microscopy showed that GPAT1, DGAT1, and lipidTOX were colocalized in the cytoplasm of WT macrophages whereas FABP4 deficiency reduced the expression of GPAT1, DGAT1, and lipidTOX in KO macrophages (Figure 5G). Quantitative analysis demonstrated that GPAT1, DGAT1, and lipidTOX levels were significantly reduced when FABP4 was absent (Figure 5H–J). Using transmission electron microscopy, we further confirmed that FABP4 deficiency compromised LA- and OA-induced lipid accumulation in macrophages (Figure 5L, Figure 5—figure supplement 1G). The role of FABP4 in lipid accumulation was corroborated by analysis of primary peritoneal macrophages from FABP4 WT and KO mice. As shown in Figure 5—figure supplement 1H and I, LA-induced neutral lipid accumulation in macrophages was significantly reduced when FABP4 was deficient. Collectively, these data suggest that FABP4 plays a pivotal role in TAG synthesis and neutral lipid accumulation in macrophages.

To further dissect the functionality of FABP4 in TAMs, we analyzed FABP4-associated gene pathways in the TCGA breast cancer datasets. Interestingly, FABP4 expression was highly associated with the lipolysis regulation pathway, including genes of β-adrenergic receptors (Adrb2), G-protein-coupled receptor signaling (Adcy4), and lipolysis enzymes (Pnpla2, Lipe) (Figure 6—figure supplement 1A–D). When FABP4 WT and KO macrophages were treated with PA or LA, FABP4 deficiency impaired LA-induced lipolysis in macrophages as evidenced by significantly reduced levels of Adrb2, Adcy4, Pnpla2, and Lipe (Figure 6A–D), suggesting a critical role of FABP4 in mediating lipolysis-associated pathways in macrophages. Given the protumor role of lipid accumulation in TAMs, we treated FABP4 WT or KO macrophages with PA, LA, or BSA, and evaluated their protumor function using trans-well tumor migration assays (Figure 6E). As shown in Figure 6F and G, compared to BSA- or PA-treated WT macrophages, LA-treated macrophages greatly promoted MDA-MB-231 cell migration. Notably, FABP4 deficiency significantly reduced LA-mediated tumor migration. Using E0771 mammary tumor cells, we confirmed the essential role of FABP4 in mediating LA-induced tumor migration effect in macrophages (Figure 6H and I). The FABP4/LA-induced tumor migration effect was not due to tumor proliferation as Ki67 expression was similar among different FA-treated groups (Figure 6—figure supplement 1E). Instead, lipid droplets formed by LA in WT macrophages were depleted following coculture with tumor cells (Figure 6J), showing the pro-migration effect of FABP4-mediated lipolysis. Furthermore, we collected the conditioned media collected from FA-treated macrophages and demonstrated that only media from LA-treated macrophages promoted migration of MDA-MB-231 and E0771 in a FABP4-dependent manner (Figure 6—figure supplement 1F–I). These data suggested that intracellular FABP4 was secreted from macrophages into the extracellular environment to mediate the pro-migration effect. Indeed, LA-treated macrophages exhibited the highest levels of FABP4 compared to other groups (Figure 6—figure supplement 1J). Altogether, FABP4 expression in macrophages was essential for LA-induced pro-tumor effects.

Recently, meta-analysis demonstrated that a high density of TAMs, especially CD163⁺ TAMs, predicted poor survival outcomes in breast cancer (Allison et al., 2023). When we analyzed pro-tumor activity of TAMs using human breast cancer tissue, we verified that high expression of the macrophage marker CD163 was significantly associated with a reduced overall survival of these breast cancer patients in multiple breast cancer datasets (Figure 7—figure supplement 1A–1D). To verify the results from these publicly accessible datasets, we collected breast cancer tissue specimens from a cohort of 59 women with different subtypes of breast cancer (Supplementary file 1) and performed immunohistochemical (IHC) staining. Interestingly, CD163⁺ TAMs were positively correlated with breast tumor size (Figure 7A and B), corroborating the pro-tumor function of CD163⁺ macrophages in breast cancer. We also noticed that compared to lean patients, obese patients had more CD163⁺ macrophages in the TME (Figure 7—figure supplement 1E and F), suggesting that dysregulated lipids in obese patients promoted macrophage tumor infiltration. Given the pivotal role of FABP4 in lipid metabolism in macrophages, we analyzed FABP4 expression in CD163⁺ TAMs. Interestingly, FABP4 expression significantly correlated with CD163⁺ macrophages (p<0.001; Figure 7C). Among the cohort, 22% of patients exhibited lymph node and distant metastasis (Figure 7D). When we compared macrophages and FABP4 expression in tumor tissues in patients with or without tumor metastasis (Figure 7E), we showed that tumors from metastatic patients had higher expression of CD163⁺ macrophages (Figure 7F) and FABP4 expression (Figure 7G). We also examined the association of other variables, including age, tumor grade, multifocality, lymphovascular invasion, mortality, hormone status, with CD163 and FABP4 expression in tumor tissues. Higher tumor grade and mortality were associated with more FABP4 expression in macrophages (Figure 7H, Supplementary file 2). Moreover, triple negative breast cancer exhibited higher FABP4 expression in macrophages than ER⁺ or HER2⁺ breast cancer (Figure 7—figure supplement 1G and H). Collectively, CD163⁺ TAMs, especially those with FABP4 expression, are positively associated with tumor growth and metastasis in breast cancer patients.

In this study, we have identified FABP4 as a crucial lipid chaperone responsible for mediating lipid accumulation and lipolysis in TAMs. However, the mechanisms by which FABP4/FA complexes secreted from TAMs interact with tumor cells and mediate tumor metastasis remain unknown. While some potential FABP4 receptors, such as membrane phospholipids (Hao et al., 2018b), desmoglein 2 (Chen et al., 2023), cytokeratin 1 (Saavedra et al., 2015), have been reported in various cell types, further investigations are needed to determine how FABP4 and its lipid ligands facilitate tumor metastasis - either by initiating oncogenic signals, by providing energy sources or both.

Fabp4^–/– mice and their wild-type (WT) mice (C57B/6 background) were maintained and bred under specific pathogen free conditions with water and regular chow ad libitum in the animal facility of the University of Iowa in accordance with the approved protocol by the University of Iowa Institutional Animal Care and Use Committee (#1042385). As breast cancer mainly occurs in female, female mice at 8–12 weeks age were used for collection of primary peritoneal macrophages, splenic macrophages and culture of BMMs.

Fresh or formalin-fixed, paraffin-embedded breast cancer tissues were collected from the Breast and Molecular Epidemiology Resources (BMER) or Department of Pathology, University of Iowa. Breast sections were cut and transported to the study team without further patient identifiers. The University of Iowa Institutional Review Board approved the study protocol (#201003791, 202107133) prior to sample acquisition.

WT and age matched Fabp4^-/- mice were euthanized by CO2 inhalation and death was confirmed via subsequent cardiac puncture. Peritoneal macrophages were isolated by injecting 10 mL PBS (Gibco) into the mouse peritoneal cavity and incubated with 3 min with periodic abdominal palpation. EtOH was used to sterilize the surgical site and the skin of the abdomen was reflected. Using a 10 mL syringe coupled with a 25 G needle (Becton Dickinson and Company), PBS from the peritoneal cavity was aspirated, moved to a 15 mL conical (Avantor) tube, centrifuged, and resuspended in RPMI media with 10% FBS. To obtain bBMMs, femurs were dissected, muscle tissue was thoroughly removed, and marrow was flushed with PBS into a 15 mL conical tube by way of a 10 mL syringe coupled with a 25 G needle. Isolated marrow cells were resuspended in red blood cell lysis buffer (Tonbo Biosciences) for 15 min on ice, washed with PBS, filtered through 40 µm cell strainers (VWR) and plated on 100 mm culture dishes (Greiner Bio-One) at a concentration of 5x10⁶ cells/mL in RPMI media with 10% FBS and 30 ng/mL recombinant mouse macrophage colony-stimulating factor (M-CSF; BioLegend). After 48 hr and again after 120 hr from initial plating cell culture media was replaced with fresh media containing M-CSF. Cells were scrutinized after 168 hr in culture.

Immortalized macrophage cell lines were established from FABP4^-/- (FABP4^−/− macrophages) or WT mice (WT macrophages) as described previously (Zhang et al., 2017c; Makowski et al., 2005; Clemons-Miller et al., 2000). Briefly, BMMs were culture as outlined above however following differentiation J2-virus conditioned media was used for 2 hr and replaced with M-CSF media. This procedure was repeated after 24 hr. Subsequently M-CSF media was removed step-wise over the course of 2 weeks at which point cells could be effectively passages. This procedure was completed for marrow cells obtained from a C57Bl/6 J mouse as well as aged match FABP4^-/- mouse. Breast cancer cell migration assays were completed using E0771 mouse mammary gland cells and MDA-MB 231 human mammary gland cells (ATCC).

Due to their insolubility, all FFAs were conjugated with bovine serum albumin (BSA) as described previously (Zhang et al., 2017c; Zhang et al., 2014; Jin et al., 2021b; Zhang et al., 2017a). Briefly, all FFAs used in the current study were purchased from Nu-Chek Prep (MN). PA (5 mM, S1109), SA (5 Mm, S-1111), OA (5 Mm, S-1120) or LA (5 Mm, S-1127) were respectively conjugated with 2 mM of endotoxin-free, FA-free BSA (Proliant, Cat# 69760) in PBS. FA-BSA conjugates were sonicated until dissolved. All FA-BSA conjugates or BSA control were filtered through a 0.2 μM sterile filter before use in cellular studies.

Predesigned duplex small interfering RNA was obtained from Integrated DNA Technologies (IDT) specific to C/EBPα and C/EBPβ. J2-Immortalized macrophages were transfected using jetPRIME (Polyplus) as instructed by the manufacture. Efficacy of RNAi was confirmed by RT-PCR for each experiment.

Cells were lysed and RNA was extracted using the PureLink RNA mini kit (Invitrogen) as per manufactures instructions. RNA concentration was determined and standardized by the aid of a photospectrometer and complementary DNA was synthesized by use of QuantiTect reverse transcription kit (QIAGEN). Power SYBR Green PCR Master Mix (Applied Biosystems) was included with target primers (Supplementary file 3) on MicroAmp optical 384-well reaction plates (Applied Biosystems) and a QuantStudio 7 Flex (Thermo Fisher) and the provided softeware (Design and Analysis Software V2.3, Thermo Fisher) was used to determine relative mRNA concentration by way of ΔΔCt method with hprt1 as an internal control. Standard PCR reaction parameters were used: 95 °C hold stage for 10 min, 40 cycles of PCR reaction of 95 °C for 15 s and 60 °C for 1 min, and a melt curve stage used 0.5 °C steps from 60°C to 65°C.

The Seahorse mitochondrial stress test (Agilent) was performed as specified by the manufacturer. Briefly, cells were plated at a concentration of 5x10⁴ per well in RPMI media with 10% FBS. Before the assay cells were wash twice with RPMI-XF media with no FBS and incubated at 37 °C for 15 min during equipment initialization. supplemented with 100 µM BSA, PA, SA, OA, or LA. The acute injection protocol was utilized where some wells (n=5 for each FA for each genotype) received injections of RPMI-XF supplemented with 200 µM BSA, PA, SA, OA, or LA, and all wells subsequently received injections of 2.5 µM oligomycin, 1 µM FCCP, and finally 1 µM Rotenone/Antimycin/Hoechst. Assays were ran using the XFe96 extracellular flux analyzer (Agilent) coupled with automated normalization based on live cell number at the conclusion of the assay.

Formalin fixed Paraffin Embedded (FFPE) tissue specimens were deparaffinized in xylene and rehydrated through exposure to graded ethanol solutions. Antigen retrieval was performed by incubating the slides in 10 mM Citrate buffer pH6.0 at 95 °C for 15 min. Then, quenching of endogenous peroxidase activity was performed by incubating the slides in BLOXALL Endogenous Blocking Solution (Vector Laboratories, SP-6000) for 10 min at room temperature. Slides were incubated with a primary antibody overnight at 4 °C, a secondary antibody for 30 min at room temperature (All the primary and secondary antibodies used in this study are summarized in key resources table). DAB peroxidase substrate (Vector Laboratories, SK-4100) or Vector Red alkaline phosphatase substrate (Vector Laboratories, SK-5100) were used for staining. Counterstain was performed by incubating the slides in Hematoxylin (Leica Gill III 3801541) for 10 s at room temperature. Slides were mounted with VectaMount Express Mounting Medium (vector laboratories, H-5700–60), and were scanned by slide scanner (Leica Aperio GT 450) for quantification analysis. Hot spot quantification of CD163 and FABP4 expression in macrophages (per 1x2 mm² area within the tumor tissue) was automatically calculated by the HALO software (Indica Labs).

Cells were fixed with 2.5% glutaraldehyde (in 0.1 M Sodium cacodylate buffer [pH 7.4]) overnight at 4 °C. samples were postfixed with 1% Osmium tetroxide for 1 hr and then rinsed in 0.1 M Sodium cacodylate buffer. Following serial alcohol dehydration (50%, 75%, 95%, and 100%), the samples were embedded in Epon 12 (Ted Pella, Redding, CA). Ultramicrotomy was performed, and ultrathin sections (70 nm) were poststained with uranyl acetate and lead citrate. Samples were examined with a Hitachi HT-7800 transmission electron microscope (TEM) (Tokyo, Japan).

Cells were fixed with 4% paraformaldehyde (PFA), and then were dehydrated in graded concentrations of ethanol and embedded in LR White resin (Electron Microscopy Sciences, 14380). Ultrathin sections (70 nm) on grids were incubated with blocking solution (Aurion, 905.003) for 1 hr, and then incubated with primary antibody overnight at 4 °C, and then incubated with secondary antibody (Aurion, Rabbit-anti-Goat IgG (H&L) Ultra Small) for 2 hr. Grids were re-fixed with 1% glutaraldehyde for 15 min and were incubated in sliver enhancement (Aurion, 500.033) for 25 min. Sections were then stained with uranyl acetate and lead citrate and examined with a Hitachi HT-7800 transmission electron microscope (TEM; Tokyo, Japan).

For each use Oil Red O powder (Alfa Aesar) was solubilized in 100% isopropanol (75 mg in 25 mL) and mixed on a orbital shaker for 30 min to make a stock solution. For a working solution Oil Red O stock solution was used at a ratio of 3 parts Oil Red O to 2 parts ddH₂O and mixed on an orbital shaker for 10 min. The working solution was filtered twice through 0.2 µm syringe filter and used immediately on 4% PFA fixed cells. During the second of three washes with ddH₂O, DAPI was included at 1:5000 for nuclear identification. Images were taken with an ECHO Revolution2 (ECHO) epifluorescence microscope and processed using ImageJ/Fiji. Following image collection, for the purpose of semi-quantitative analysis, samples were washed with 60% isopropanol three times for 5 min each and the residual Oil Red O was solubilized with 100% Isopropanol and absorbance was assessed by spectrophotometer at 492 nm, 100% isopropanol was used as a background.

The following anti-mouse conjugated antibodies were used to label macrophages surface receptors from peritoneal and bone marrow derived cells: PE/Cyanine7-F4/80, BV711-CD11c, BV605-I-A/I-E (BioLegend), BUV737-CD11b, and BUV563-CD45 (BD Biosciences). Unconjugated antibodies were used for quantification of intracellular markers: m/rFABP4 (R&D Systems), GPAM (Invitrogen), DGAT1 (Sigma-Aldrich), C/EBPα (Cell Signaling), C/EBPβ (DSHB). The following organelle dyes were used: BODIPY 493/503 (Lipid Droplets; Molecular Probes), Cell Navigator Live Cell Endoplasmic Reticulum Staining Kit Red (ATT Bioquest), and LysoTracker Blue DND-22 (Invitrogen). Dead cells were identified and removed from analysis via Ghost Dye Violet 510 (Tonbo Biosciences). Briefly, live cells were scrutinized unless fixation was required for intracellular targets in which case 4% PFA was used for 30 min to fix the cells and 0.2% Triton X-100 in PBS was used for permeabilization. Antibodies or organelle dyes were incubated with cells for 30 min on ice in the dark. When necessary secondary antibodies were incubated with cells for 1 hr on ice in the dark. Data was acquired using a Cytek Aurora 5-laser flow cytometer or ImageStream cytometer (CytekBio) and subsequently analyzed FlowJo V10 (BD Biosciences).

For the migration assay, we treated WT and FABP4 KO macrophages with BSA, PA, or LA (100 μM) for 4 hr in a 24-well plate. After washing them three times with PBS, we seeded E0771 or MDA-MB 231 cells (5×104) into transwell inserts with a pore size of 8 μm (Corning). These cells were incubated with the FA-treated macrophages, respectively, to allow tumor cells to migrate. After a 24 hr culture at 37 °C, we removed the tumor cells inside the transwells that had not migrated using cotton swabs. The tumor cells that had migrated to the outside of the insert membrane were fixed with 4% PFA for 20 min and then stained with 0.1% crystal violet for 10 min. We counted the numbers of migrated E0771 and MDA-MB 231 cells in 12 random high fields under a microscope. Each assay was performed in triplicate wells. Additionally, in some experiments, we used conditional media (RPMI-1640 without FBS) collected from BSA or FA-treated WT and KO macrophages to coculture with tumor cells and observe medium-induced tumor cell migration.

Macrophages from mouse spleen were used for the 10 x Genomics 3’ expression single cell assay. Briefly, splenocytes were isolated from two pairs of age-matched WT and FABP4-/- female mice. Dead cells were excluded using Zombie-Violet and anti-CD16/32 antibody was used to block Fc receptors before additional staining with Alexa Fluor 488-conjugated anti-mouse F4/480 antibody and APC-conjugated anti-mouse CD11b antibody. The F4/80+/CD11b+/Zombie-Voilet- cells were sorted using a FACS Aira III instrument and resuspended at a concentration of 1000 cell/μL cold PBS with 0.04% non-acetylated BSA. An equal number of 5000 targeted cells from each sample were prepared to create GEMs. The GEM generation/barcoding, post GEM-reverse transcription cDNA amplification, and library construction were performed according to the manufacturer’s guidelines.

Single-cell gene counting was performed by Cell Ranger (10 X Genomics, version 7.1.0) using the refdata-gex-mm10-2020-A reference transcriptome. The resulting data matrices were subsequently imported into R (version 4.2.3) and analyzed using the Seurat package (version 4.9.9). Cells with less than 200 features or with percent mitochondrial gene expression greater than 5% were excluded from the analysis. The data across samples was integrated using the Integrate Data function. Gene expression was normalized and scaled using the default parameters. Based on visual inspection of the elbow plot, the first 30 Principal Components (PC) were used in UMAP-based dimensional reduction. The FindClusters function, with a resolution of 0.5, was then used to assign cells to clusters. The FindMarkers function was used to identify genes differentially expressed between wild-type and knockout samples or between different clusters within the same sample type. Feature plots and violin plots were used to visualize the results.

FABP4 excreted from cultured cells were analyzed with CircuLex Mouse FABP4 ELISA Kit (Medical and Biological Laboratories) as instructed by the manufactures. Briefly, for the ELISA 100 µL of conditioned media was incubated in a 96-well plate precoated with FABP4 antibody, washed four times, incubated with the supplied secondary antibody, washed, and incubated with the developer to finally analyze with a spectrophotometer using a ratio of the absorbance at 450 nm and 550 nm.

All data were presented as the mean ± SD unless notified specifically. All experiment as indicated were performed by at least three independent experiments or technical replicates. For in vitro experiments, a two-tailed, unpaired student t-test, two-way ANOVA followed by Bonferroni’s multiple comparison test, were performed by GraphPad Prism 9. For in vivo experiments analyzing associations of FABP4, TAMs and other factors in breast cancer patients, the Kruskal-Wallis test was applied to compare whether the outcome variable is significantly different across the different levels of a categorical predictor variable. Multiple linear regression models were employed to examine the association between the predictor variables and the outcome, where log transformation and Box Cox transformation were used to meet the underlying model assumption for normality. All statistical analyses were carried out in the statistical analysis software R. A statistics test is claimed to be significant if the p-value is less than 0.05.